1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3 * net/core/fib_rules.c Generic Routing Rules
4 *
5 * Authors: Thomas Graf <tgraf@suug.ch>
6 */
7
8 #include <linux/types.h>
9 #include <linux/kernel.h>
10 #include <linux/slab.h>
11 #include <linux/list.h>
12 #include <linux/module.h>
13 #include <net/net_namespace.h>
14 #include <net/sock.h>
15 #include <net/fib_rules.h>
16 #include <net/ip_tunnels.h>
17
18 static const struct fib_kuid_range fib_kuid_range_unset = {
19 KUIDT_INIT(0),
20 KUIDT_INIT(~0),
21 };
22
23 bool fib_rule_matchall(const struct fib_rule *rule)
24 {
25 if (rule->iifindex || rule->oifindex || rule->mark || rule->tun_id ||
26 rule->flags)
27 return false;
28 if (rule->suppress_ifgroup != -1 || rule->suppress_prefixlen != -1)
29 return false;
30 if (!uid_eq(rule->uid_range.start, fib_kuid_range_unset.start) ||
31 !uid_eq(rule->uid_range.end, fib_kuid_range_unset.end))
32 return false;
33 if (fib_rule_port_range_set(&rule->sport_range))
34 return false;
35 if (fib_rule_port_range_set(&rule->dport_range))
36 return false;
37 return true;
38 }
39 EXPORT_SYMBOL_GPL(fib_rule_matchall);
40
41 int fib_default_rule_add(struct fib_rules_ops *ops,
42 u32 pref, u32 table, u32 flags)
43 {
44 struct fib_rule *r;
45
46 r = kzalloc(ops->rule_size, GFP_KERNEL);
47 if (r == NULL)
48 return -ENOMEM;
49
50 refcount_set(&r->refcnt, 1);
51 r->action = FR_ACT_TO_TBL;
52 r->pref = pref;
53 r->table = table;
54 r->flags = flags;
55 r->proto = RTPROT_KERNEL;
56 r->fr_net = ops->fro_net;
57 r->uid_range = fib_kuid_range_unset;
58
59 r->suppress_prefixlen = -1;
60 r->suppress_ifgroup = -1;
61
62 /* The lock is not required here, the list in unreacheable
63 * at the moment this function is called */
64 list_add_tail(&r->list, &ops->rules_list);
65 return 0;
66 }
67 EXPORT_SYMBOL(fib_default_rule_add);
68
69 static u32 fib_default_rule_pref(struct fib_rules_ops *ops)
70 {
71 struct list_head *pos;
72 struct fib_rule *rule;
73
74 if (!list_empty(&ops->rules_list)) {
75 pos = ops->rules_list.next;
76 if (pos->next != &ops->rules_list) {
77 rule = list_entry(pos->next, struct fib_rule, list);
78 if (rule->pref)
79 return rule->pref - 1;
80 }
81 }
82
83 return 0;
84 }
85
86 static void notify_rule_change(int event, struct fib_rule *rule,
87 struct fib_rules_ops *ops, struct nlmsghdr *nlh,
88 u32 pid);
89
90 static struct fib_rules_ops *lookup_rules_ops(struct net *net, int family)
91 {
92 struct fib_rules_ops *ops;
93
94 rcu_read_lock();
95 list_for_each_entry_rcu(ops, &net->rules_ops, list) {
96 if (ops->family == family) {
97 if (!try_module_get(ops->owner))
98 ops = NULL;
99 rcu_read_unlock();
100 return ops;
101 }
102 }
103 rcu_read_unlock();
104
105 return NULL;
106 }
107
108 static void rules_ops_put(struct fib_rules_ops *ops)
109 {
110 if (ops)
111 module_put(ops->owner);
112 }
113
114 static void flush_route_cache(struct fib_rules_ops *ops)
115 {
116 if (ops->flush_cache)
117 ops->flush_cache(ops);
118 }
119
120 static int __fib_rules_register(struct fib_rules_ops *ops)
121 {
122 int err = -EEXIST;
123 struct fib_rules_ops *o;
124 struct net *net;
125
126 net = ops->fro_net;
127
128 if (ops->rule_size < sizeof(struct fib_rule))
129 return -EINVAL;
130
131 if (ops->match == NULL || ops->configure == NULL ||
132 ops->compare == NULL || ops->fill == NULL ||
133 ops->action == NULL)
134 return -EINVAL;
135
136 spin_lock(&net->rules_mod_lock);
137 list_for_each_entry(o, &net->rules_ops, list)
138 if (ops->family == o->family)
139 goto errout;
140
141 list_add_tail_rcu(&ops->list, &net->rules_ops);
142 err = 0;
143 errout:
144 spin_unlock(&net->rules_mod_lock);
145
146 return err;
147 }
148
149 struct fib_rules_ops *
150 fib_rules_register(const struct fib_rules_ops *tmpl, struct net *net)
151 {
152 struct fib_rules_ops *ops;
153 int err;
154
155 ops = kmemdup(tmpl, sizeof(*ops), GFP_KERNEL);
156 if (ops == NULL)
157 return ERR_PTR(-ENOMEM);
158
159 INIT_LIST_HEAD(&ops->rules_list);
160 ops->fro_net = net;
161
162 err = __fib_rules_register(ops);
163 if (err) {
164 kfree(ops);
165 ops = ERR_PTR(err);
166 }
167
168 return ops;
169 }
170 EXPORT_SYMBOL_GPL(fib_rules_register);
171
172 static void fib_rules_cleanup_ops(struct fib_rules_ops *ops)
173 {
174 struct fib_rule *rule, *tmp;
175
176 list_for_each_entry_safe(rule, tmp, &ops->rules_list, list) {
177 list_del_rcu(&rule->list);
178 if (ops->delete)
179 ops->delete(rule);
180 fib_rule_put(rule);
181 }
182 }
183
184 void fib_rules_unregister(struct fib_rules_ops *ops)
185 {
186 struct net *net = ops->fro_net;
187
188 spin_lock(&net->rules_mod_lock);
189 list_del_rcu(&ops->list);
190 spin_unlock(&net->rules_mod_lock);
191
192 fib_rules_cleanup_ops(ops);
193 kfree_rcu(ops, rcu);
194 }
195 EXPORT_SYMBOL_GPL(fib_rules_unregister);
196
197 static int uid_range_set(struct fib_kuid_range *range)
198 {
199 return uid_valid(range->start) && uid_valid(range->end);
200 }
201
202 static struct fib_kuid_range nla_get_kuid_range(struct nlattr **tb)
203 {
204 struct fib_rule_uid_range *in;
205 struct fib_kuid_range out;
206
207 in = (struct fib_rule_uid_range *)nla_data(tb[FRA_UID_RANGE]);
208
209 out.start = make_kuid(current_user_ns(), in->start);
210 out.end = make_kuid(current_user_ns(), in->end);
211
212 return out;
213 }
214
215 static int nla_put_uid_range(struct sk_buff *skb, struct fib_kuid_range *range)
216 {
217 struct fib_rule_uid_range out = {
218 from_kuid_munged(current_user_ns(), range->start),
219 from_kuid_munged(current_user_ns(), range->end)
220 };
221
222 return nla_put(skb, FRA_UID_RANGE, sizeof(out), &out);
223 }
224
225 static int nla_get_port_range(struct nlattr *pattr,
226 struct fib_rule_port_range *port_range)
227 {
228 const struct fib_rule_port_range *pr = nla_data(pattr);
229
230 if (!fib_rule_port_range_valid(pr))
231 return -EINVAL;
232
233 port_range->start = pr->start;
234 port_range->end = pr->end;
235
236 return 0;
237 }
238
239 static int nla_put_port_range(struct sk_buff *skb, int attrtype,
240 struct fib_rule_port_range *range)
241 {
242 return nla_put(skb, attrtype, sizeof(*range), range);
243 }
244
245 static int fib_rule_match(struct fib_rule *rule, struct fib_rules_ops *ops,
246 struct flowi *fl, int flags,
247 struct fib_lookup_arg *arg)
248 {
249 int ret = 0;
250
251 if (rule->iifindex && (rule->iifindex != fl->flowi_iif))
252 goto out;
253
254 if (rule->oifindex && (rule->oifindex != fl->flowi_oif))
255 goto out;
256
257 if ((rule->mark ^ fl->flowi_mark) & rule->mark_mask)
258 goto out;
259
260 if (rule->tun_id && (rule->tun_id != fl->flowi_tun_key.tun_id))
261 goto out;
262
263 if (rule->l3mdev && !l3mdev_fib_rule_match(rule->fr_net, fl, arg))
264 goto out;
265
266 if (uid_lt(fl->flowi_uid, rule->uid_range.start) ||
267 uid_gt(fl->flowi_uid, rule->uid_range.end))
268 goto out;
269
270 ret = ops->match(rule, fl, flags);
271 out:
272 return (rule->flags & FIB_RULE_INVERT) ? !ret : ret;
273 }
274
275 int fib_rules_lookup(struct fib_rules_ops *ops, struct flowi *fl,
276 int flags, struct fib_lookup_arg *arg)
277 {
278 struct fib_rule *rule;
279 int err;
280
281 rcu_read_lock();
282
283 list_for_each_entry_rcu(rule, &ops->rules_list, list) {
284 jumped:
285 if (!fib_rule_match(rule, ops, fl, flags, arg))
286 continue;
287
288 if (rule->action == FR_ACT_GOTO) {
289 struct fib_rule *target;
290
291 target = rcu_dereference(rule->ctarget);
292 if (target == NULL) {
293 continue;
294 } else {
295 rule = target;
296 goto jumped;
297 }
298 } else if (rule->action == FR_ACT_NOP)
299 continue;
300 else
301 err = ops->action(rule, fl, flags, arg);
302
303 if (!err && ops->suppress && ops->suppress(rule, arg))
304 continue;
305
306 if (err != -EAGAIN) {
307 if ((arg->flags & FIB_LOOKUP_NOREF) ||
308 likely(refcount_inc_not_zero(&rule->refcnt))) {
309 arg->rule = rule;
310 goto out;
311 }
312 break;
313 }
314 }
315
316 err = -ESRCH;
317 out:
318 rcu_read_unlock();
319
320 return err;
321 }
322 EXPORT_SYMBOL_GPL(fib_rules_lookup);
323
324 static int call_fib_rule_notifier(struct notifier_block *nb,
325 enum fib_event_type event_type,
326 struct fib_rule *rule, int family,
327 struct netlink_ext_ack *extack)
328 {
329 struct fib_rule_notifier_info info = {
330 .info.family = family,
331 .info.extack = extack,
332 .rule = rule,
333 };
334
335 return call_fib_notifier(nb, event_type, &info.info);
336 }
337
338 static int call_fib_rule_notifiers(struct net *net,
339 enum fib_event_type event_type,
340 struct fib_rule *rule,
341 struct fib_rules_ops *ops,
342 struct netlink_ext_ack *extack)
343 {
344 struct fib_rule_notifier_info info = {
345 .info.family = ops->family,
346 .info.extack = extack,
347 .rule = rule,
348 };
349
350 ops->fib_rules_seq++;
351 return call_fib_notifiers(net, event_type, &info.info);
352 }
353
354 /* Called with rcu_read_lock() */
355 int fib_rules_dump(struct net *net, struct notifier_block *nb, int family,
356 struct netlink_ext_ack *extack)
357 {
358 struct fib_rules_ops *ops;
359 struct fib_rule *rule;
360 int err = 0;
361
362 ops = lookup_rules_ops(net, family);
363 if (!ops)
364 return -EAFNOSUPPORT;
365 list_for_each_entry_rcu(rule, &ops->rules_list, list) {
366 err = call_fib_rule_notifier(nb, FIB_EVENT_RULE_ADD,
367 rule, family, extack);
368 if (err)
369 break;
370 }
371 rules_ops_put(ops);
372
373 return err;
374 }
375 EXPORT_SYMBOL_GPL(fib_rules_dump);
376
377 unsigned int fib_rules_seq_read(struct net *net, int family)
378 {
379 unsigned int fib_rules_seq;
380 struct fib_rules_ops *ops;
381
382 ASSERT_RTNL();
383
384 ops = lookup_rules_ops(net, family);
385 if (!ops)
386 return 0;
387 fib_rules_seq = ops->fib_rules_seq;
388 rules_ops_put(ops);
389
390 return fib_rules_seq;
391 }
392 EXPORT_SYMBOL_GPL(fib_rules_seq_read);
393
394 static struct fib_rule *rule_find(struct fib_rules_ops *ops,
395 struct fib_rule_hdr *frh,
396 struct nlattr **tb,
397 struct fib_rule *rule,
398 bool user_priority)
399 {
400 struct fib_rule *r;
401
402 list_for_each_entry(r, &ops->rules_list, list) {
403 if (rule->action && r->action != rule->action)
404 continue;
405
406 if (rule->table && r->table != rule->table)
407 continue;
408
409 if (user_priority && r->pref != rule->pref)
410 continue;
411
412 if (rule->iifname[0] &&
413 memcmp(r->iifname, rule->iifname, IFNAMSIZ))
414 continue;
415
416 if (rule->oifname[0] &&
417 memcmp(r->oifname, rule->oifname, IFNAMSIZ))
418 continue;
419
420 if (rule->mark && r->mark != rule->mark)
421 continue;
422
423 if (rule->suppress_ifgroup != -1 &&
424 r->suppress_ifgroup != rule->suppress_ifgroup)
425 continue;
426
427 if (rule->suppress_prefixlen != -1 &&
428 r->suppress_prefixlen != rule->suppress_prefixlen)
429 continue;
430
431 if (rule->mark_mask && r->mark_mask != rule->mark_mask)
432 continue;
433
434 if (rule->tun_id && r->tun_id != rule->tun_id)
435 continue;
436
437 if (r->fr_net != rule->fr_net)
438 continue;
439
440 if (rule->l3mdev && r->l3mdev != rule->l3mdev)
441 continue;
442
443 if (uid_range_set(&rule->uid_range) &&
444 (!uid_eq(r->uid_range.start, rule->uid_range.start) ||
445 !uid_eq(r->uid_range.end, rule->uid_range.end)))
446 continue;
447
448 if (rule->ip_proto && r->ip_proto != rule->ip_proto)
449 continue;
450
451 if (rule->proto && r->proto != rule->proto)
452 continue;
453
454 if (fib_rule_port_range_set(&rule->sport_range) &&
455 !fib_rule_port_range_compare(&r->sport_range,
456 &rule->sport_range))
457 continue;
458
459 if (fib_rule_port_range_set(&rule->dport_range) &&
460 !fib_rule_port_range_compare(&r->dport_range,
461 &rule->dport_range))
462 continue;
463
464 if (!ops->compare(r, frh, tb))
465 continue;
466 return r;
467 }
468
469 return NULL;
470 }
471
472 #ifdef CONFIG_NET_L3_MASTER_DEV
473 static int fib_nl2rule_l3mdev(struct nlattr *nla, struct fib_rule *nlrule,
474 struct netlink_ext_ack *extack)
475 {
476 nlrule->l3mdev = nla_get_u8(nla);
477 if (nlrule->l3mdev != 1) {
478 NL_SET_ERR_MSG(extack, "Invalid l3mdev attribute");
479 return -1;
480 }
481
482 return 0;
483 }
484 #else
485 static int fib_nl2rule_l3mdev(struct nlattr *nla, struct fib_rule *nlrule,
486 struct netlink_ext_ack *extack)
487 {
488 NL_SET_ERR_MSG(extack, "l3mdev support is not enabled in kernel");
489 return -1;
490 }
491 #endif
492
493 static int fib_nl2rule(struct sk_buff *skb, struct nlmsghdr *nlh,
494 struct netlink_ext_ack *extack,
495 struct fib_rules_ops *ops,
496 struct nlattr *tb[],
497 struct fib_rule **rule,
498 bool *user_priority)
499 {
500 struct net *net = sock_net(skb->sk);
501 struct fib_rule_hdr *frh = nlmsg_data(nlh);
502 struct fib_rule *nlrule = NULL;
503 int err = -EINVAL;
504
505 if (frh->src_len)
506 if (!tb[FRA_SRC] ||
507 frh->src_len > (ops->addr_size * 8) ||
508 nla_len(tb[FRA_SRC]) != ops->addr_size) {
509 NL_SET_ERR_MSG(extack, "Invalid source address");
510 goto errout;
511 }
512
513 if (frh->dst_len)
514 if (!tb[FRA_DST] ||
515 frh->dst_len > (ops->addr_size * 8) ||
516 nla_len(tb[FRA_DST]) != ops->addr_size) {
517 NL_SET_ERR_MSG(extack, "Invalid dst address");
518 goto errout;
519 }
520
521 nlrule = kzalloc(ops->rule_size, GFP_KERNEL);
522 if (!nlrule) {
523 err = -ENOMEM;
524 goto errout;
525 }
526 refcount_set(&nlrule->refcnt, 1);
527 nlrule->fr_net = net;
528
529 if (tb[FRA_PRIORITY]) {
530 nlrule->pref = nla_get_u32(tb[FRA_PRIORITY]);
531 *user_priority = true;
532 } else {
533 nlrule->pref = fib_default_rule_pref(ops);
534 }
535
536 nlrule->proto = tb[FRA_PROTOCOL] ?
537 nla_get_u8(tb[FRA_PROTOCOL]) : RTPROT_UNSPEC;
538
539 if (tb[FRA_IIFNAME]) {
540 struct net_device *dev;
541
542 nlrule->iifindex = -1;
543 nla_strlcpy(nlrule->iifname, tb[FRA_IIFNAME], IFNAMSIZ);
544 dev = __dev_get_by_name(net, nlrule->iifname);
545 if (dev)
546 nlrule->iifindex = dev->ifindex;
547 }
548
549 if (tb[FRA_OIFNAME]) {
550 struct net_device *dev;
551
552 nlrule->oifindex = -1;
553 nla_strlcpy(nlrule->oifname, tb[FRA_OIFNAME], IFNAMSIZ);
554 dev = __dev_get_by_name(net, nlrule->oifname);
555 if (dev)
556 nlrule->oifindex = dev->ifindex;
557 }
558
559 if (tb[FRA_FWMARK]) {
560 nlrule->mark = nla_get_u32(tb[FRA_FWMARK]);
561 if (nlrule->mark)
562 /* compatibility: if the mark value is non-zero all bits
563 * are compared unless a mask is explicitly specified.
564 */
565 nlrule->mark_mask = 0xFFFFFFFF;
566 }
567
568 if (tb[FRA_FWMASK])
569 nlrule->mark_mask = nla_get_u32(tb[FRA_FWMASK]);
570
571 if (tb[FRA_TUN_ID])
572 nlrule->tun_id = nla_get_be64(tb[FRA_TUN_ID]);
573
574 err = -EINVAL;
575 if (tb[FRA_L3MDEV] &&
576 fib_nl2rule_l3mdev(tb[FRA_L3MDEV], nlrule, extack) < 0)
577 goto errout_free;
578
579 nlrule->action = frh->action;
580 nlrule->flags = frh->flags;
581 nlrule->table = frh_get_table(frh, tb);
582 if (tb[FRA_SUPPRESS_PREFIXLEN])
583 nlrule->suppress_prefixlen = nla_get_u32(tb[FRA_SUPPRESS_PREFIXLEN]);
584 else
585 nlrule->suppress_prefixlen = -1;
586
587 if (tb[FRA_SUPPRESS_IFGROUP])
588 nlrule->suppress_ifgroup = nla_get_u32(tb[FRA_SUPPRESS_IFGROUP]);
589 else
590 nlrule->suppress_ifgroup = -1;
591
592 if (tb[FRA_GOTO]) {
593 if (nlrule->action != FR_ACT_GOTO) {
594 NL_SET_ERR_MSG(extack, "Unexpected goto");
595 goto errout_free;
596 }
597
598 nlrule->target = nla_get_u32(tb[FRA_GOTO]);
599 /* Backward jumps are prohibited to avoid endless loops */
600 if (nlrule->target <= nlrule->pref) {
601 NL_SET_ERR_MSG(extack, "Backward goto not supported");
602 goto errout_free;
603 }
604 } else if (nlrule->action == FR_ACT_GOTO) {
605 NL_SET_ERR_MSG(extack, "Missing goto target for action goto");
606 goto errout_free;
607 }
608
609 if (nlrule->l3mdev && nlrule->table) {
610 NL_SET_ERR_MSG(extack, "l3mdev and table are mutually exclusive");
611 goto errout_free;
612 }
613
614 if (tb[FRA_UID_RANGE]) {
615 if (current_user_ns() != net->user_ns) {
616 err = -EPERM;
617 NL_SET_ERR_MSG(extack, "No permission to set uid");
618 goto errout_free;
619 }
620
621 nlrule->uid_range = nla_get_kuid_range(tb);
622
623 if (!uid_range_set(&nlrule->uid_range) ||
624 !uid_lte(nlrule->uid_range.start, nlrule->uid_range.end)) {
625 NL_SET_ERR_MSG(extack, "Invalid uid range");
626 goto errout_free;
627 }
628 } else {
629 nlrule->uid_range = fib_kuid_range_unset;
630 }
631
632 if (tb[FRA_IP_PROTO])
633 nlrule->ip_proto = nla_get_u8(tb[FRA_IP_PROTO]);
634
635 if (tb[FRA_SPORT_RANGE]) {
636 err = nla_get_port_range(tb[FRA_SPORT_RANGE],
637 &nlrule->sport_range);
638 if (err) {
639 NL_SET_ERR_MSG(extack, "Invalid sport range");
640 goto errout_free;
641 }
642 }
643
644 if (tb[FRA_DPORT_RANGE]) {
645 err = nla_get_port_range(tb[FRA_DPORT_RANGE],
646 &nlrule->dport_range);
647 if (err) {
648 NL_SET_ERR_MSG(extack, "Invalid dport range");
649 goto errout_free;
650 }
651 }
652
653 *rule = nlrule;
654
655 return 0;
656
657 errout_free:
658 kfree(nlrule);
659 errout:
660 return err;
661 }
662
663 static int rule_exists(struct fib_rules_ops *ops, struct fib_rule_hdr *frh,
664 struct nlattr **tb, struct fib_rule *rule)
665 {
666 struct fib_rule *r;
667
668 list_for_each_entry(r, &ops->rules_list, list) {
669 if (r->action != rule->action)
670 continue;
671
672 if (r->table != rule->table)
673 continue;
674
675 if (r->pref != rule->pref)
676 continue;
677
678 if (memcmp(r->iifname, rule->iifname, IFNAMSIZ))
679 continue;
680
681 if (memcmp(r->oifname, rule->oifname, IFNAMSIZ))
682 continue;
683
684 if (r->mark != rule->mark)
685 continue;
686
687 if (r->suppress_ifgroup != rule->suppress_ifgroup)
688 continue;
689
690 if (r->suppress_prefixlen != rule->suppress_prefixlen)
691 continue;
692
693 if (r->mark_mask != rule->mark_mask)
694 continue;
695
696 if (r->tun_id != rule->tun_id)
697 continue;
698
699 if (r->fr_net != rule->fr_net)
700 continue;
701
702 if (r->l3mdev != rule->l3mdev)
703 continue;
704
705 if (!uid_eq(r->uid_range.start, rule->uid_range.start) ||
706 !uid_eq(r->uid_range.end, rule->uid_range.end))
707 continue;
708
709 if (r->ip_proto != rule->ip_proto)
710 continue;
711
712 if (r->proto != rule->proto)
713 continue;
714
715 if (!fib_rule_port_range_compare(&r->sport_range,
716 &rule->sport_range))
717 continue;
718
719 if (!fib_rule_port_range_compare(&r->dport_range,
720 &rule->dport_range))
721 continue;
722
723 if (!ops->compare(r, frh, tb))
724 continue;
725 return 1;
726 }
727 return 0;
728 }
729
730 int fib_nl_newrule(struct sk_buff *skb, struct nlmsghdr *nlh,
731 struct netlink_ext_ack *extack)
732 {
733 struct net *net = sock_net(skb->sk);
734 struct fib_rule_hdr *frh = nlmsg_data(nlh);
735 struct fib_rules_ops *ops = NULL;
736 struct fib_rule *rule = NULL, *r, *last = NULL;
737 struct nlattr *tb[FRA_MAX + 1];
738 int err = -EINVAL, unresolved = 0;
739 bool user_priority = false;
740
741 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*frh))) {
742 NL_SET_ERR_MSG(extack, "Invalid msg length");
743 goto errout;
744 }
745
746 ops = lookup_rules_ops(net, frh->family);
747 if (!ops) {
748 err = -EAFNOSUPPORT;
749 NL_SET_ERR_MSG(extack, "Rule family not supported");
750 goto errout;
751 }
752
753 err = nlmsg_parse_deprecated(nlh, sizeof(*frh), tb, FRA_MAX,
754 ops->policy, extack);
755 if (err < 0) {
756 NL_SET_ERR_MSG(extack, "Error parsing msg");
757 goto errout;
758 }
759
760 err = fib_nl2rule(skb, nlh, extack, ops, tb, &rule, &user_priority);
761 if (err)
762 goto errout;
763
764 if ((nlh->nlmsg_flags & NLM_F_EXCL) &&
765 rule_exists(ops, frh, tb, rule)) {
766 err = -EEXIST;
767 goto errout_free;
768 }
769
770 err = ops->configure(rule, skb, frh, tb, extack);
771 if (err < 0)
772 goto errout_free;
773
774 err = call_fib_rule_notifiers(net, FIB_EVENT_RULE_ADD, rule, ops,
775 extack);
776 if (err < 0)
777 goto errout_free;
778
779 list_for_each_entry(r, &ops->rules_list, list) {
780 if (r->pref == rule->target) {
781 RCU_INIT_POINTER(rule->ctarget, r);
782 break;
783 }
784 }
785
786 if (rcu_dereference_protected(rule->ctarget, 1) == NULL)
787 unresolved = 1;
788
789 list_for_each_entry(r, &ops->rules_list, list) {
790 if (r->pref > rule->pref)
791 break;
792 last = r;
793 }
794
795 if (last)
796 list_add_rcu(&rule->list, &last->list);
797 else
798 list_add_rcu(&rule->list, &ops->rules_list);
799
800 if (ops->unresolved_rules) {
801 /*
802 * There are unresolved goto rules in the list, check if
803 * any of them are pointing to this new rule.
804 */
805 list_for_each_entry(r, &ops->rules_list, list) {
806 if (r->action == FR_ACT_GOTO &&
807 r->target == rule->pref &&
808 rtnl_dereference(r->ctarget) == NULL) {
809 rcu_assign_pointer(r->ctarget, rule);
810 if (--ops->unresolved_rules == 0)
811 break;
812 }
813 }
814 }
815
816 if (rule->action == FR_ACT_GOTO)
817 ops->nr_goto_rules++;
818
819 if (unresolved)
820 ops->unresolved_rules++;
821
822 if (rule->tun_id)
823 ip_tunnel_need_metadata();
824
825 notify_rule_change(RTM_NEWRULE, rule, ops, nlh, NETLINK_CB(skb).portid);
826 flush_route_cache(ops);
827 rules_ops_put(ops);
828 return 0;
829
830 errout_free:
831 kfree(rule);
832 errout:
833 rules_ops_put(ops);
834 return err;
835 }
836 EXPORT_SYMBOL_GPL(fib_nl_newrule);
837
838 int fib_nl_delrule(struct sk_buff *skb, struct nlmsghdr *nlh,
839 struct netlink_ext_ack *extack)
840 {
841 struct net *net = sock_net(skb->sk);
842 struct fib_rule_hdr *frh = nlmsg_data(nlh);
843 struct fib_rules_ops *ops = NULL;
844 struct fib_rule *rule = NULL, *r, *nlrule = NULL;
845 struct nlattr *tb[FRA_MAX+1];
846 int err = -EINVAL;
847 bool user_priority = false;
848
849 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*frh))) {
850 NL_SET_ERR_MSG(extack, "Invalid msg length");
851 goto errout;
852 }
853
854 ops = lookup_rules_ops(net, frh->family);
855 if (ops == NULL) {
856 err = -EAFNOSUPPORT;
857 NL_SET_ERR_MSG(extack, "Rule family not supported");
858 goto errout;
859 }
860
861 err = nlmsg_parse_deprecated(nlh, sizeof(*frh), tb, FRA_MAX,
862 ops->policy, extack);
863 if (err < 0) {
864 NL_SET_ERR_MSG(extack, "Error parsing msg");
865 goto errout;
866 }
867
868 err = fib_nl2rule(skb, nlh, extack, ops, tb, &nlrule, &user_priority);
869 if (err)
870 goto errout;
871
872 rule = rule_find(ops, frh, tb, nlrule, user_priority);
873 if (!rule) {
874 err = -ENOENT;
875 goto errout;
876 }
877
878 if (rule->flags & FIB_RULE_PERMANENT) {
879 err = -EPERM;
880 goto errout;
881 }
882
883 if (ops->delete) {
884 err = ops->delete(rule);
885 if (err)
886 goto errout;
887 }
888
889 if (rule->tun_id)
890 ip_tunnel_unneed_metadata();
891
892 list_del_rcu(&rule->list);
893
894 if (rule->action == FR_ACT_GOTO) {
895 ops->nr_goto_rules--;
896 if (rtnl_dereference(rule->ctarget) == NULL)
897 ops->unresolved_rules--;
898 }
899
900 /*
901 * Check if this rule is a target to any of them. If so,
902 * adjust to the next one with the same preference or
903 * disable them. As this operation is eventually very
904 * expensive, it is only performed if goto rules, except
905 * current if it is goto rule, have actually been added.
906 */
907 if (ops->nr_goto_rules > 0) {
908 struct fib_rule *n;
909
910 n = list_next_entry(rule, list);
911 if (&n->list == &ops->rules_list || n->pref != rule->pref)
912 n = NULL;
913 list_for_each_entry(r, &ops->rules_list, list) {
914 if (rtnl_dereference(r->ctarget) != rule)
915 continue;
916 rcu_assign_pointer(r->ctarget, n);
917 if (!n)
918 ops->unresolved_rules++;
919 }
920 }
921
922 call_fib_rule_notifiers(net, FIB_EVENT_RULE_DEL, rule, ops,
923 NULL);
924 notify_rule_change(RTM_DELRULE, rule, ops, nlh,
925 NETLINK_CB(skb).portid);
926 fib_rule_put(rule);
927 flush_route_cache(ops);
928 rules_ops_put(ops);
929 kfree(nlrule);
930 return 0;
931
932 errout:
933 kfree(nlrule);
934 rules_ops_put(ops);
935 return err;
936 }
937 EXPORT_SYMBOL_GPL(fib_nl_delrule);
938
939 static inline size_t fib_rule_nlmsg_size(struct fib_rules_ops *ops,
940 struct fib_rule *rule)
941 {
942 size_t payload = NLMSG_ALIGN(sizeof(struct fib_rule_hdr))
943 + nla_total_size(IFNAMSIZ) /* FRA_IIFNAME */
944 + nla_total_size(IFNAMSIZ) /* FRA_OIFNAME */
945 + nla_total_size(4) /* FRA_PRIORITY */
946 + nla_total_size(4) /* FRA_TABLE */
947 + nla_total_size(4) /* FRA_SUPPRESS_PREFIXLEN */
948 + nla_total_size(4) /* FRA_SUPPRESS_IFGROUP */
949 + nla_total_size(4) /* FRA_FWMARK */
950 + nla_total_size(4) /* FRA_FWMASK */
951 + nla_total_size_64bit(8) /* FRA_TUN_ID */
952 + nla_total_size(sizeof(struct fib_kuid_range))
953 + nla_total_size(1) /* FRA_PROTOCOL */
954 + nla_total_size(1) /* FRA_IP_PROTO */
955 + nla_total_size(sizeof(struct fib_rule_port_range)) /* FRA_SPORT_RANGE */
956 + nla_total_size(sizeof(struct fib_rule_port_range)); /* FRA_DPORT_RANGE */
957
958 if (ops->nlmsg_payload)
959 payload += ops->nlmsg_payload(rule);
960
961 return payload;
962 }
963
964 static int fib_nl_fill_rule(struct sk_buff *skb, struct fib_rule *rule,
965 u32 pid, u32 seq, int type, int flags,
966 struct fib_rules_ops *ops)
967 {
968 struct nlmsghdr *nlh;
969 struct fib_rule_hdr *frh;
970
971 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*frh), flags);
972 if (nlh == NULL)
973 return -EMSGSIZE;
974
975 frh = nlmsg_data(nlh);
976 frh->family = ops->family;
977 frh->table = rule->table < 256 ? rule->table : RT_TABLE_COMPAT;
978 if (nla_put_u32(skb, FRA_TABLE, rule->table))
979 goto nla_put_failure;
980 if (nla_put_u32(skb, FRA_SUPPRESS_PREFIXLEN, rule->suppress_prefixlen))
981 goto nla_put_failure;
982 frh->res1 = 0;
983 frh->res2 = 0;
984 frh->action = rule->action;
985 frh->flags = rule->flags;
986
987 if (nla_put_u8(skb, FRA_PROTOCOL, rule->proto))
988 goto nla_put_failure;
989
990 if (rule->action == FR_ACT_GOTO &&
991 rcu_access_pointer(rule->ctarget) == NULL)
992 frh->flags |= FIB_RULE_UNRESOLVED;
993
994 if (rule->iifname[0]) {
995 if (nla_put_string(skb, FRA_IIFNAME, rule->iifname))
996 goto nla_put_failure;
997 if (rule->iifindex == -1)
998 frh->flags |= FIB_RULE_IIF_DETACHED;
999 }
1000
1001 if (rule->oifname[0]) {
1002 if (nla_put_string(skb, FRA_OIFNAME, rule->oifname))
1003 goto nla_put_failure;
1004 if (rule->oifindex == -1)
1005 frh->flags |= FIB_RULE_OIF_DETACHED;
1006 }
1007
1008 if ((rule->pref &&
1009 nla_put_u32(skb, FRA_PRIORITY, rule->pref)) ||
1010 (rule->mark &&
1011 nla_put_u32(skb, FRA_FWMARK, rule->mark)) ||
1012 ((rule->mark_mask || rule->mark) &&
1013 nla_put_u32(skb, FRA_FWMASK, rule->mark_mask)) ||
1014 (rule->target &&
1015 nla_put_u32(skb, FRA_GOTO, rule->target)) ||
1016 (rule->tun_id &&
1017 nla_put_be64(skb, FRA_TUN_ID, rule->tun_id, FRA_PAD)) ||
1018 (rule->l3mdev &&
1019 nla_put_u8(skb, FRA_L3MDEV, rule->l3mdev)) ||
1020 (uid_range_set(&rule->uid_range) &&
1021 nla_put_uid_range(skb, &rule->uid_range)) ||
1022 (fib_rule_port_range_set(&rule->sport_range) &&
1023 nla_put_port_range(skb, FRA_SPORT_RANGE, &rule->sport_range)) ||
1024 (fib_rule_port_range_set(&rule->dport_range) &&
1025 nla_put_port_range(skb, FRA_DPORT_RANGE, &rule->dport_range)) ||
1026 (rule->ip_proto && nla_put_u8(skb, FRA_IP_PROTO, rule->ip_proto)))
1027 goto nla_put_failure;
1028
1029 if (rule->suppress_ifgroup != -1) {
1030 if (nla_put_u32(skb, FRA_SUPPRESS_IFGROUP, rule->suppress_ifgroup))
1031 goto nla_put_failure;
1032 }
1033
1034 if (ops->fill(rule, skb, frh) < 0)
1035 goto nla_put_failure;
1036
1037 nlmsg_end(skb, nlh);
1038 return 0;
1039
1040 nla_put_failure:
1041 nlmsg_cancel(skb, nlh);
1042 return -EMSGSIZE;
1043 }
1044
1045 static int dump_rules(struct sk_buff *skb, struct netlink_callback *cb,
1046 struct fib_rules_ops *ops)
1047 {
1048 int idx = 0;
1049 struct fib_rule *rule;
1050 int err = 0;
1051
1052 rcu_read_lock();
1053 list_for_each_entry_rcu(rule, &ops->rules_list, list) {
1054 if (idx < cb->args[1])
1055 goto skip;
1056
1057 err = fib_nl_fill_rule(skb, rule, NETLINK_CB(cb->skb).portid,
1058 cb->nlh->nlmsg_seq, RTM_NEWRULE,
1059 NLM_F_MULTI, ops);
1060 if (err)
1061 break;
1062 skip:
1063 idx++;
1064 }
1065 rcu_read_unlock();
1066 cb->args[1] = idx;
1067 rules_ops_put(ops);
1068
1069 return err;
1070 }
1071
1072 static int fib_valid_dumprule_req(const struct nlmsghdr *nlh,
1073 struct netlink_ext_ack *extack)
1074 {
1075 struct fib_rule_hdr *frh;
1076
1077 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*frh))) {
1078 NL_SET_ERR_MSG(extack, "Invalid header for fib rule dump request");
1079 return -EINVAL;
1080 }
1081
1082 frh = nlmsg_data(nlh);
1083 if (frh->dst_len || frh->src_len || frh->tos || frh->table ||
1084 frh->res1 || frh->res2 || frh->action || frh->flags) {
1085 NL_SET_ERR_MSG(extack,
1086 "Invalid values in header for fib rule dump request");
1087 return -EINVAL;
1088 }
1089
1090 if (nlmsg_attrlen(nlh, sizeof(*frh))) {
1091 NL_SET_ERR_MSG(extack, "Invalid data after header in fib rule dump request");
1092 return -EINVAL;
1093 }
1094
1095 return 0;
1096 }
1097
1098 static int fib_nl_dumprule(struct sk_buff *skb, struct netlink_callback *cb)
1099 {
1100 const struct nlmsghdr *nlh = cb->nlh;
1101 struct net *net = sock_net(skb->sk);
1102 struct fib_rules_ops *ops;
1103 int idx = 0, family;
1104
1105 if (cb->strict_check) {
1106 int err = fib_valid_dumprule_req(nlh, cb->extack);
1107
1108 if (err < 0)
1109 return err;
1110 }
1111
1112 family = rtnl_msg_family(nlh);
1113 if (family != AF_UNSPEC) {
1114 /* Protocol specific dump request */
1115 ops = lookup_rules_ops(net, family);
1116 if (ops == NULL)
1117 return -EAFNOSUPPORT;
1118
1119 dump_rules(skb, cb, ops);
1120
1121 return skb->len;
1122 }
1123
1124 rcu_read_lock();
1125 list_for_each_entry_rcu(ops, &net->rules_ops, list) {
1126 if (idx < cb->args[0] || !try_module_get(ops->owner))
1127 goto skip;
1128
1129 if (dump_rules(skb, cb, ops) < 0)
1130 break;
1131
1132 cb->args[1] = 0;
1133 skip:
1134 idx++;
1135 }
1136 rcu_read_unlock();
1137 cb->args[0] = idx;
1138
1139 return skb->len;
1140 }
1141
1142 static void notify_rule_change(int event, struct fib_rule *rule,
1143 struct fib_rules_ops *ops, struct nlmsghdr *nlh,
1144 u32 pid)
1145 {
1146 struct net *net;
1147 struct sk_buff *skb;
1148 int err = -ENOBUFS;
1149
1150 net = ops->fro_net;
1151 skb = nlmsg_new(fib_rule_nlmsg_size(ops, rule), GFP_KERNEL);
1152 if (skb == NULL)
1153 goto errout;
1154
1155 err = fib_nl_fill_rule(skb, rule, pid, nlh->nlmsg_seq, event, 0, ops);
1156 if (err < 0) {
1157 /* -EMSGSIZE implies BUG in fib_rule_nlmsg_size() */
1158 WARN_ON(err == -EMSGSIZE);
1159 kfree_skb(skb);
1160 goto errout;
1161 }
1162
1163 rtnl_notify(skb, net, pid, ops->nlgroup, nlh, GFP_KERNEL);
1164 return;
1165 errout:
1166 if (err < 0)
1167 rtnl_set_sk_err(net, ops->nlgroup, err);
1168 }
1169
1170 static void attach_rules(struct list_head *rules, struct net_device *dev)
1171 {
1172 struct fib_rule *rule;
1173
1174 list_for_each_entry(rule, rules, list) {
1175 if (rule->iifindex == -1 &&
1176 strcmp(dev->name, rule->iifname) == 0)
1177 rule->iifindex = dev->ifindex;
1178 if (rule->oifindex == -1 &&
1179 strcmp(dev->name, rule->oifname) == 0)
1180 rule->oifindex = dev->ifindex;
1181 }
1182 }
1183
1184 static void detach_rules(struct list_head *rules, struct net_device *dev)
1185 {
1186 struct fib_rule *rule;
1187
1188 list_for_each_entry(rule, rules, list) {
1189 if (rule->iifindex == dev->ifindex)
1190 rule->iifindex = -1;
1191 if (rule->oifindex == dev->ifindex)
1192 rule->oifindex = -1;
1193 }
1194 }
1195
1196
1197 static int fib_rules_event(struct notifier_block *this, unsigned long event,
1198 void *ptr)
1199 {
1200 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1201 struct net *net = dev_net(dev);
1202 struct fib_rules_ops *ops;
1203
1204 ASSERT_RTNL();
1205
1206 switch (event) {
1207 case NETDEV_REGISTER:
1208 list_for_each_entry(ops, &net->rules_ops, list)
1209 attach_rules(&ops->rules_list, dev);
1210 break;
1211
1212 case NETDEV_CHANGENAME:
1213 list_for_each_entry(ops, &net->rules_ops, list) {
1214 detach_rules(&ops->rules_list, dev);
1215 attach_rules(&ops->rules_list, dev);
1216 }
1217 break;
1218
1219 case NETDEV_UNREGISTER:
1220 list_for_each_entry(ops, &net->rules_ops, list)
1221 detach_rules(&ops->rules_list, dev);
1222 break;
1223 }
1224
1225 return NOTIFY_DONE;
1226 }
1227
1228 static struct notifier_block fib_rules_notifier = {
1229 .notifier_call = fib_rules_event,
1230 };
1231
1232 static int __net_init fib_rules_net_init(struct net *net)
1233 {
1234 INIT_LIST_HEAD(&net->rules_ops);
1235 spin_lock_init(&net->rules_mod_lock);
1236 return 0;
1237 }
1238
1239 static void __net_exit fib_rules_net_exit(struct net *net)
1240 {
1241 WARN_ON_ONCE(!list_empty(&net->rules_ops));
1242 }
1243
1244 static struct pernet_operations fib_rules_net_ops = {
1245 .init = fib_rules_net_init,
1246 .exit = fib_rules_net_exit,
1247 };
1248
1249 static int __init fib_rules_init(void)
1250 {
1251 int err;
1252 rtnl_register(PF_UNSPEC, RTM_NEWRULE, fib_nl_newrule, NULL, 0);
1253 rtnl_register(PF_UNSPEC, RTM_DELRULE, fib_nl_delrule, NULL, 0);
1254 rtnl_register(PF_UNSPEC, RTM_GETRULE, NULL, fib_nl_dumprule, 0);
1255
1256 err = register_pernet_subsys(&fib_rules_net_ops);
1257 if (err < 0)
1258 goto fail;
1259
1260 err = register_netdevice_notifier(&fib_rules_notifier);
1261 if (err < 0)
1262 goto fail_unregister;
1263
1264 return 0;
1265
1266 fail_unregister:
1267 unregister_pernet_subsys(&fib_rules_net_ops);
1268 fail:
1269 rtnl_unregister(PF_UNSPEC, RTM_NEWRULE);
1270 rtnl_unregister(PF_UNSPEC, RTM_DELRULE);
1271 rtnl_unregister(PF_UNSPEC, RTM_GETRULE);
1272 return err;
1273 }
1274
1275 subsys_initcall(fib_rules_init);
1276
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.